Radiomics Analysis: Complemental But Unessential for Routine Clinical Practice in Predicting Microvascular Invasion of Hepatocellular Carcinoma

Background: Microvascular invasion (MVI) is a prognostic factor of patients with hepatocellular carcinoma (HCC). Preoperative prediction of MVI is challenging. Use of Radiomics as clinical biomarker necessitates amelioration. We therefore estimated whether Radiomics can provide complementary value for predicting MVI risk in HCC patients as compared with a liver imaging score (LI-score) system derived from radiologists' interpretation. Methods: 495 patients (300 in training, 50 in validation and 145 in testing data set) with histologically confirmed HCC between Jan 2009 and Aug 2017 were retrospectively reviewed. All the patients have intact preoperative clinicoradiological records. 93 radiomic features were extracted from each patient and radiomics signature were built by using recursive feature selection support vector machine (RFE-SVM) method. Odds-ratio (OR) regression model was performed to determine risk factors of MVI and construct the predictive nomogram. A decision curve analysis was applied to teste the incremental value of Radiomics score (RS) to LI-score for predicting MVI. Findings: 9 radiomics features were selected to build a radiomics signature that was significantly associated with MVI (p < 0.001). An MVI risk estimation nomogram, which incorporated higher aspartate aminotransferase (AST), higher α-fetoprotein (AFP), non-smooth tumor margin, extrahepatic growth pattern, ill-defined pseudo-capsule, peritumoral arterial enhancement, positive radio-genomic venous invasion (RVI) score and higher RS, was established with an area under curve (AUC) of 0.894 (95%CI, 0.864-0.920). Adding RS to LI-score scheme did not achieve significant improvement in predictive performance and did not add additional net benefits in a decision curve analysis. Interpretation: Radiomics is still a complemental tool that offer limited additional benefit to conventional radiologic method for predicting MVI risk of HCC. The use of Radiomics as clinical biomarkers still needs further study. Funding: China Postdoctoral Fund. A Key Social Development Program for the Ministry of Science and Technology of Jiangsu Province. Declaration of Interest: The authors have declared no conflicts of interest. Ethical Approval: This retrospective study involved routine at a single medical institution. Ethics committee approval was granted by local institutional ethics review board with a waiver of written informed consent. All procedures performed in studies involving human participants were in accordance with the 1964 Helsinki declaration and its later amendments.